“The bottom-cost place to place AI will probably be in house, and that will probably be true inside two years, perhaps three on the newest,” SpaceX founder Elon Musk advised the World Financial Discussion board in Davos this previous January, as his firm was preparing to go public.
Later that month, SpaceX filed an utility with the Federal Communications Fee for an orbital information middle constellation of as much as 1 million satellites in low Earth orbit, 500 to 2,000 kilometers above Earth. And simply three days earlier than the IPO, he mentioned some preliminary design specs for a brand new AI-1 satellite data center in a video interview.
Musk is susceptible to hyperbole in relation to timelines. Full self-driving cars by 2017. First human mission to Mars in 2024. Ten thousand Optimus humanoid robots by the end of 2025. Et cetera. For orbital data centers, which he says will probably be a cheap different to terrestrial information facilities inside three years, the mathematics gained’t make sense for a number of years, if ever.
Think about this: There are roughly 14,500 active satellites in orbit. Musk’s Starlink constellation accounts for about two thirds of those. Each the launch cadences and satellite-manufacturing capability must scale up astronomically to deploy 1,000,000 orbital information middle satellites.
For context, there have been roughly 7,000 orbital launches in all of human history. To loft 1 million satellites into low Earth orbit on SpaceX’s Starship, which is designed to hold as much as 60 satellites per automobile, would require 16,666 launches completely dedicated to satellite tv for pc deployments. Contemplating that SpaceX launched a report 165 orbital missions in 2025, even at 10 instances that cadence, it might take a decade. And the way lengthy would it not take to construct 1 million satellites, given Starlink’s current pace of around 4,000 per year and a beneficiant tenfold enhance in capability? In need of a producing revolution, attempt 25 years.
The fact is that the imaginative and prescient of huge constellations of orbital information facilities is nowhere near being realized.
As this month’s cowl story, “Why Orbital Data Centers Are So Hard” by Andrew Cavalier of ABI Research, makes clear, the fact is that the imaginative and prescient of huge constellations of orbital information facilities is nowhere near being realized.
Dina Genkina, IEEE Spectrum’s computing and {hardware} editor, put the concept into perspective: “Starcloud (a startup that has utilized to the FCC for an 88,000 orbital information middle satellite tv for pc constellation) sent one Nvidia H100 GPU in space so far. Their radiator was too weak to let the chip run at full energy.”
As Cavalier exhibits, cooling even a single Nvidia H100 GPU in house is troublesome: It attracts 700 watts, which would require 1.4 sq. meters of radiator at 60 °C. A 40-kilowatt rack of servers will want an 80-m² radiator; a 100-megawatt information middle would require 2,500 of these radiators. Some astronomers are understandably involved that 1,000,000 satellites with big radiative wings would blot out the celebrities.
So if the economics doesn’t make sense, if the chips are on the mercy of the radiative ravages of house, and if humanity will lose its view of the celebrities, to not point out growing the chance of triggering the Kessler syndrome, why are the hyperscalers hyping orbital information facilities?
Genkina supplied the apparent reply: candy, candy moolah. “The Elon Musk a part of it’s truthfully genius as a result of he’s received xAI constructing the info facilities, SpaceX sending them to house, and Tesla constructing photo voltaic panels,” Genkina says. “It’s virtually like he’s paying himself.”
From Your Web site Articles
Associated Articles Across the Internet
